Age | Commit message (Collapse) | Author |
|
|
|
|
|
|
|
|
|
- Only pass `torrent.infoHash` to the Chunk Store constructor, instead
of the `Torrent`
instance itself, to prevent accidental memory leaks of the `Torrent`
object by the
store. (Open an issue if you were using other properties. They can be
re-added.)
- Non-fatal errors with a single torrent will be emitted at
`torrent.on('error')`. You
should listen to this event. Previously, all torrent errors were also
emitted on
`client.on('error')` and handling `torrent.on('error')` was optional.
This design is
better since now it is possible to distinguish between fatal client
errors
(`client.on('error')`) when the whole client becomes unusable versus
recoverable errors
where only a single torrent fails (`torrent.on('error')`) but the
client can continue to
be used. However, if there is no `torrent.on('error')` event, then
the error will be
forwarded to `client.on('error')`. This prevents crashing the client
when the user
only has a listener on the client, but it makes it impossible for
them to determine
a client error versus a torrent error.
- Errors creating a torrent with `client.seed` are now emitted on the
returned `torrent`
object instead of the client (unless there is no event listeners on
`torrent.on('error')` as previously discussed). The torrent object is
now also destroyed
automatically for the user, as was probably expected.
- If `client.get` is passed a `Torrent` instance, it now only returns
it if it is present
in the client.
|
|
### Added
- `client.listening` property to signal whether TCP server is listening
for incoming
connections.
### Changed
- Merged `Swarm` class into `Torrent` object. Properties on
`torrent.swarm` (like
`torrent.swarm.wires`) now exist on `torrent` (e.g. `torrent.wires`).
- `torrent.addPeer` can no longer be called before the `infoHash` event
has been
emitted.
- Remove `torrent.on('listening')` event. Use `client.on('listening')`
instead.
- Remove support from `TCPPool` for listening on multiple ports. This
was not used by
WebTorrent and just added complexity. There is now a single `TCPPool`
instance for the
whole WebTorrent client.
- Deprecate: Do not use `client.download()` anymore. Use `client.add()`
instead.
- Deprecate: Do not use `torrent.swarm` anymore. Use `torrent` instead.
### Fixed
- When there is a `torrent.on('error')` listener, don't also emit
`client.on('error')`.
- Do not return existing torrent object when duplicate torrent is
added. Fire an
`'error'` event instead.
- Memory leak of `Torrent` object caused by `RarityMap`
- Memory leak of `Torrent` object caused by `TCPPool`
- `client.ratio` and `torrent.ratio` are now calculated as `uploaded /
received` instead
of `uploaded / downloaded`.
|
|
|
|
|
|
When I started the WebTorrent project I thought there were going to
need to be two separate client implementations (bittorrent-client and
webtorrent-client) that would get tied together in a higher-level
module.
Fortunately, this was not necessary because of the awesome “browser”
field support in browserify. By substituting just a few modules, we can
make the same module (webtorrent) work in node AND the browser, with
the same codebase!
So, from now on, you can just `require(‘webtorrent’)` in node or the
browser, and it will just work. You can also `npm install webtorrent`
if you want to use bittorrent in a node app or script. Lastly, you can
`npm install webtorrent -g` if you want to use webtorrent as a command
line app (it installs a `webtorrent` command).
|